To reduce mother-to-child transmission of human immunodeficiency virus (HIV) in resource-poor settings, the World Health Organization recommends exclusive breast-feeding for 6 months, followed by rapid weaning if replacement feeding is affordable, feasible, available, safe, and sustainable. In the Kisumu Breastfeeding Study (trial registration: Clinicaltrials.gov identifier NCT00146380), infants of HIV-infected mothers who received antiretroviral therapy experienced high rates of diarrhea at weaning. To address this problem, mothers in the Kisumu Breastfeeding Study were given safe water storage vessels, hygiene education, and bleach for household water treatment. We compared the incidence of diarrhea in infants enrolled before (cohort A) and after (cohort B) implementation of the intervention. Cohort B infants experienced less diarrhea than cohort A infants, before and after weaning (P < .001 and P = .047, respectively); however, during the weaning period, there were no differences in the frequency of diarrhea between cohorts (P = 0.89). Testing of stored water in cohort B homes indicated high adherence (monthly range, 80%-95%) to recommended chlorination practices. Among infants who were weaned early, provision of safe water may be insufficient to prevent weaning-associated diarrhea.